It’s no secret that we’re not all the same when it comes to our kitchen skills, or our propensity to spend time in the kitchen. This is one of the key drivers driving the use of new AI developments to shape the future of home and commercial kitchens. (And the obvious windfall you’ll get from it.)
In the latest development, from the University of Cambridge’s engineering department, researchers taught an AI-programmed robot “chef” to make its own salad using the machine’s brain after being instructed in an instructional video.
By watching a human slowly make eight different salads, and using mathematical formulas to allow machines to translate visual cues, the robot can identify the ingredients and create a variety from the mind’s “cookbook”. I was able to prepare a nice salad. Additionally, the robot accumulated knowledge as it progressed and was able to come up with a ninth recipe on its own. However, this original recipe seems to have simply rearranged the ingredients used in another salad.
“We wanted to know if we could train robot chefs in the same step-by-step way that humans do, by identifying ingredients and how they combine in a dish.” said lead author of the paper, Grzegorz Soczakki of the Cambridge School of Engineering. .
Recipes were limited to five ingredients: broccoli, carrots, apples, bananas, and oranges. Three recipes were coleslaw variations, three were fruit salad varieties, and two were vegetable dishes. It’s safe to say that this robot chef isn’t quite ready for wedding catering, or even date night. Of her 16 instructional videos watched by the robot, the human chef’s demonstration identified the correct recipe 93% of the time, even though it could only detect him 83% of the time.
Cambridge University
Of course, there were obvious limits to the robot’s capabilities. Not only could it not perform the complex movements commonly required of human chefs in food preparation, neural networks trained on the Microsoft Common Objects in Context (COCO) image recognition dataset were still very rudimentary. .
But the researchers were still impressed by how the robot chef was able to infer behavior from the video. For example, if a demonstrator held a knife in one hand and a carrot in the other, the robot recognized that the vegetables had to be passed through a slicer during cooking.
“Our robots aren’t interested in the kind of food videos that go viral on social media, they’re just too hard to follow,” Soczakki said. “But maybe these robot chefs will get better at identifying ingredients in cooking videos faster, allowing them to learn every recipe using sites like YouTube.”
Automating food production is certainly not a new idea. It has given us consumer gadgets like CookingPal and, most recently, culinary monstrosities like 3D-printed “cakes.” Machine learning has the potential to be game-changing for both home chefs and professional kitchens at some point, but it’s fair to say that no line cook is serious about it yet.
Cambridge University
“These recipes are not complicated. “It’s amazing how many nuances the robot can detect,” he added.
The study was published in a journal IEEE access.
While it is no chef Watch the audition tape and see Cambridge’s Robot Chef in action in this video.
Robot ‘chef’ learns to recreate recipes by watching cooking videos
Source: University of Cambridge
